Research map / Point clouds / Point cloud classification
TARS: Traffic-Aware Radar Scene Flow Estimation
Jialong Wu, Marco Braun, Dominic Spata, Matthias Rottmann
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add traffic-level understanding on top of point-level motion (ablation) | TARS · Proprietary radar dataset | MEPE (m): 0.178 -> 0.069 | Improved |
| add OD feature map to the TVF encoder (ablation) | TARS · Proprietary radar dataset | AccR (%): 63.3 -> 86.8 | Improved |
| use coarse 2m TVF grid instead of 1m fine grid (ablation) | TARS · Proprietary radar dataset | MEPE (m): 0.104 -> 0.069 | Improved |
| use global attention in TVF encoder instead of local convolutions (ablation) | TARS · Proprietary radar dataset | AccS (%): 65.6 -> 69.8 | Improved |
| use global attention instead of local convolution (ablation) | TARS · Proprietary radar dataset | SEPE (m): 0.031 -> 0.038 | Hurt |
| attend to 9 TVF cells instead of 5 (ablation) | TARS · Proprietary radar dataset | AccS (%): 67.2 -> 69.8 | Improved |
| expand attention to 13 TVF cells instead of 9 (ablation) | TARS · Proprietary radar dataset | AccS (%): 69.8 -> 59.9 | Hurt |
| add PointGRU temporal module to PointPWC-Net | PointPWC-Net · Proprietary radar dataset | MEPE (m): 0.453 -> 0.213 | Improved |
| add background loss L_bg (ablation) | TARS-ego · VOD (radar scene flow) | EPE (m): 0.107 -> 0.092 | Improved |
| raise lambda_bg from 0.5 to 1.00 (ablation) | TARS-no-ego · VOD (radar scene flow) | MRNE (m): 0.065 -> 0.067 | Hurt |
| raise lambda_bg from 0.5 to 1.00 (ablation) | TARS-no-ego · VOD (radar scene flow) | AccR (%): 59.3 -> 65.8 | Improved |
| vary N_TVF among 5, 9, 13 (ablation) | TARS-ego · VOD (radar scene flow) | EPE (m): 0.094 -> 0.092 -> 0.093 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SLIM — SLIM: Self-Supervised LiDAR Scene Flow and Motion Segmentation (ICCV 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ICP-Flow: LiDAR Scene Flow Estimation with ICP (CVPR 2024 · similarity 0.89)
- UniFlow: Zero-Shot LiDAR Scene Flow for Autonomous Driving (ECCV 2026 · similarity 0.89)
- RaLiFlow: Scene Flow Estimation with 4D Radar and LiDAR Point Clouds (AAAI 2026 · similarity 0.88)
- Weakly Supervised Cross-Modal Learning for 4D Radar Scene Flow Estimation (ICML 2026 · similarity 0.88)
- Motion Cues from Image-based Point Tracking for LiDAR Scene Flow Estimation (NeurIPS 2026 · similarity 0.87)
- RadarMOSEVE: A Spatial-Temporal Transformer Network for Radar-Only Moving Object Segmentation and Ego-Velocity Estimation (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
